eval_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -3.100578546524048,
    "eval_logits/rejected": -3.094856023788452,
    "eval_logps/chosen": -60.72254180908203,
    "eval_logps/rejected": -66.06378173828125,
    "eval_loss": 0.6891194581985474,
    "eval_rewards/accuracies": 0.5910780429840088,
    "eval_rewards/chosen": -0.020106395706534386,
    "eval_rewards/margins": 0.008730227127671242,
    "eval_rewards/rejected": -0.028836622834205627,
    "eval_runtime": 383.2975,
    "eval_samples": 4304,
    "eval_samples_per_second": 11.229,
    "eval_steps_per_second": 1.404
}